Contribution
The paper turns a philosophers’ intuition into something measured: that “the pain is in my fingertip; the fingertip is in my mouth; therefore the pain is in my mouth” fails. It shows that lay participants share the intuition. They share it as much for tissue damage and inflammation, where neither of the two standard explanations applies. It offers a third explanation, a misleading implicature in the conclusion, and gives two pieces of evidence for it: a correlation across items, and a wording manipulation.
Key insight
The failure of the pain-in-mouth argument is not special to pain. Any “X is in my mouth” that suggests something is wrong with the mouth is rejected when X is really in a finger that is in the mouth. So the argument cannot show that “pain” is intensional or that its “in” is not spatial. Read at face value, “there is a pain in my fingertip” locates the pain in the fingertip.
Assumptions
- What “fails” means. The authors take the target to be “the ordinary understanding of the premises and conclusion” that drives the intuition (p. 4), not validity in a technical sense. So agreement with the conclusion on a 7-point scale is the measure.
- Analogy by matched behaviour. If pain, tissue damage and inflammation are rejected “to a roughly equal degree” (p. 6), one explanation should cover all three. This is a parsimony argument (“Occam’s razor”, p. 8), not a demonstration that the same process operates in each case.
- Misleadingness is measurable by a direct question. A separate group rated how misleading “[item] in their mouth” would be if one later learned the item was in or on a thumb that was in the mouth (p. 7).
- No bodily-versus-mental contrast term. There is no item that everyone treats as mental and that still takes a bodily “in”. The test is pain against bodily states and a paper-clip, not pain against mental states.
Key results
- Study 1 (pp. 5–6, Fig. 1). 174 participants on Amazon Mechanical Turk (54.0% women, mean age 36.6, fn 5) each judged one argument. Mean agreement with the conclusion: pain 2.43, tissue damage 2.53, inflammation 2.86, paper-clip (“on my fingertip”) 6.10. The first three are below the midpoint (pain t(43) = −5.44, p < 0.001); the paper-clip is above it (t(40) = 10.038). Pain did not differ from tissue damage (p = .812) or inflammation (p = 0.308); Wilcoxon tests agreed (fn 6). A reviewer’s item, serum in a fingertip, drew 5.93 from 37 participants (fn 7). The authors grant that the results “do not necessarily show that the pain-in-mouth argument does not suffer from the problems that Tye and Noordhof identify” (p. 6).
- Study 2 (pp. 6–8, Fig. 2). Fourteen items in or on a finger. Part one: 70 participants rated misleadingness of all fourteen, within subjects. Part two: 560 new participants, 40 per item, rated “There is [item] in Susan’s mouth”, recruited through the Philosophical Personality website. Part two replicated Study 1: pain 2.25, tissue damage 2.55, inflammation 3.40, paper-clip 5.25 (fn 10). Regressing mean acceptance on mean misleadingness: all items F(1,12) = 104.83, R² = 0.897; “in” items only F(1,5) = 11.79, p = 0.019, R² = 0.702 (fn 11). On the rendered Figure 2, pain is the lowest point on both axes, with tissue damage, cut, inflammation, bone, sore, splinter and infection near it, and paper-clip, perfume, nail, gasoline and blood on at the top. The authors warn that “one should be wary of moving from correlation to causation” (p. 8).
- Study 3 (pp. 8–9). 244 MTurk participants, randomised to the original conclusion or one prefixed “Although it might sound misleading to say that there is a pain in Susan’s mouth, technically speaking there is a pain located in her mouth”, for the four lowest-rated items. Revised against original: pain 3.79 against 2.10, inflammation 4.14 against 2.42, cut 4.97 against 2.66, tissue damage 4.66 against 3.21. The main effect of wording was F(1, 236) = 48.71, η² = .17, and there was no wording × item interaction (F = 0.54, p = 0.659). Under the revised wording, only cut was significantly above the midpoint; pain and inflammation were not (fn 13).
- The conclusion (p. 9). “Not only can the pain-in-mouth argument no longer be used to support a representationalist view of pain, the results suggest that pain is a state of the body, not of the mind. The semantics of reports like ‘there is pain in my fingertip,’ taken at face value, tells us that pains are located in body parts.”
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Lay people reject the conclusion of the pain-in-mouth argument | moderate | Study 1 and the Study 2 replication, two samples, about 40 per cell |
| C2 | They reject it about as much when “pain” is replaced by tissue damage or inflammation, and accept it for a paper-clip or serum | moderate | Study 1, no significant pain–bodily-term differences; pairwise nulls without equivalence tests |
| C3 | Acceptance across items tracks rated misleadingness of the conclusion | moderate as a correlation | Study 2, 14 item means, R² 0.897; between-group, item-level, not causal |
| C4 | Cancelling the implicature raises acceptance | moderate | Study 3, η² .17, no interaction; pain rises only to the midpoint |
| C5 | Implicature, not an intensional context or a non-spatial “in”, explains the argument’s failure | moderate as best explanation | C2–C4 plus parsimony; the authors grant that the other factors may also operate |
| C6 | The pain-in-mouth argument no longer supports representationalism about pain | moderate, conditional on C5 | argument from C5 |
| C7 | Pain is a state of the body, not of the mind | weak | the authors’ closing inference from C5; no item contrasts pain with a mental-state term |
Concepts
- Pain-in-mouth argument — Block’s (1983) example of an apparently invalid inference with “in”, used by Tye and Noordhof.
- Intensional context — Tye’s diagnosis: “pain”, like “want”, blocks substitution, as it would if pains were representational states.
- State-attributing “in” — Noordhof’s diagnosis, as in “a hole in my shoe”.
- Conversational implicature — what the conclusion suggests beyond what it says (here, that something is wrong with the mouth), by Grice’s maxims of relevance and manner.
- Bodily conception of pain — the authors’ programme’s view that the folk locate pains in body parts, not in the mind.
Connections
The paper continues Sytsma and Reuter’s experimental work on the folk concept of pain (Sytsma 2010; Reuter 2011; Reuter, Phillips & Sytsma 2014; Sytsma & Reuter 2017), cited and not read here. Footnote 3 places it in a “growing series of results” that “many of the commonsense facts that philosophers have claimed about phenomenal consciousness are not in fact commonsense”, citing Sytsma & Machery 2010. Its targets are Tye’s representationalism about pain and Noordhof’s reply, both in Analysis.
Bearing on the record
- The rubber hand THEORYs: no bearing. THEORY-766, THEORY-769, THEORY-767 and THEORY-768 are about what synchronous stimulation does to reports and to ownership. This paper studies how people judge a sentence about where a pain is. It meets no promotion test, partly or wholly, and none of its results is a refutation condition of any of them. Nothing in it concerns ownership or agency, so it does not bear on THEORY-458 either.
- Felt location and referred touch. THEORY-766’s referred-touch clause concerns reports that touch “seemed” to be felt where the rubber hand was touched. That is a report of where a sensation seems to be, by someone having it. This paper is about how a third party reads a locative sentence about a pain whose location is not in doubt. The only connection is a point of language that I am adding, not a result: if “in” stays spatial for sensations, as the authors argue, a “feeling the touch … in the location where I saw the rubber hand touched” item asks a spatial question. That says nothing about whether the answer is driven by the stimulation or by expectation, which is what THEORY-766 leaves open. No bearing on its promotion test.
- THEORY-767 and interoception. Neither “interoception” nor any bodily signal is studied. No bearing.
- QUESTION-048. This is evidence of a different kind from any the question lists. It is not a report of an experience, nor a no-report, deficit or architectural measure. It is a judgement about a sentence, made by people who were not in pain at the time. What it can show is how “pain” behaves in ordinary inference: it patterns with “tissue damage” and “inflammation” in this one schema. It cannot show where pain is, or whether pain is mental, and C7 takes that step. The authors’ own wording is conditional (“if our proposal is correct”, p. 9). The record should hold C2–C4 and leave C7 to the authors.
- Representationalism (LIT-635, NOTE-511). Block’s C3 is that representationists “have not said what the content of orgasm or pain experience is”. Tye’s reply in this literature is that feelings of pain represent tissue damage (fn 4). This paper attacks a supporting argument for that view, the intensional reading of “pain”, and not the view itself. C6 takes away one consideration for representationalism. It is not evidence against it.
- Problem intuitions (LIT-425, NOTE-354). Chalmers’s taxonomy includes intuitions he takes to be widely shared, and his footnote 7 absorbs Sytsma & Machery’s robot-feels-pain result. This paper belongs to the programme Chalmers was answering. It does not test a problem intuition, but its footnote 3 claims a pattern in which philosophers’ “commonsense facts” about consciousness are not common sense.
- Affect (THEORY-045, QUESTION-059). THEORY-045 counts pain among felt affect, as the experience of the body’s regulatory state. A folk concept that puts pain in body parts would fit that account and so would one that puts it in the mind, because THEORY-045 is a claim about what is felt, not about what “pain” means. No bearing, and none on QUESTION-059.
- Self, persons (TERM-045, THEORY-417, ADR-024). None.
- No instruction for machine-learning practice.
Limitations
- Small cells, online samples. About 40 per cell in Studies 1 and 2. The nulls between pain and the bodily terms carry the analogy, and they have no equivalence test. One sample came from a philosophy personality-test website.
- Item-level correlation. C3 regresses fourteen means rated by two different groups. It cannot show that any one participant rejects the conclusion because it is misleading, and the authors say so.
- The misleadingness question builds in the answer. It asks how misleading the claim would be “if you were later told” the item was on a thumb in the mouth. Rating that misleading is close to rejecting the inference, so the two measures may share a construal and not only a cause.
- The Study 3 wording is leading. “Technically speaking there is a pain located in her mouth” asserts the conclusion it then asks about. Raised agreement fits cancelled implicature, and also fits deference to the stated answer. Pain reached only 3.79, not above the midpoint.
- No mental comparison term. Without an item that people treat as mental and that still takes “in”, the studies cannot show that “pain” differs from a mental-state term in this schema. That is the contrast C7 needs.
- Version. Read from the accepted manuscript; the typeset text was not compared.
Open questions
- Would a sensation term that people call mental (“an itch”, “a tingle”, “a feeling of pressure”) pattern with pain or with the paper-clip?
- Does the result hold for people in pain, or for chronic and diffuse pain, which Borg, Harrison, Stazicker & Salomons (2019) say the bodily view’s studies leave out?
- Does the implicature account survive a measure of misleadingness that does not mention the thumb-in-mouth outcome?